Issue 61 - Article 14

Path-dependency culture in humanitarian decision-making: why it was hard to change direction in Haiti

May 30, 2014
Kate Crawford, Jim Kennedy and Alison Killing
A cluster coordination meeting in Port-au-Prince, Haiti, February 2010

This article questions whether the humanitarian cluster system’s mechanisms for strategic thinking really allow plans to be adapted as situations change, with a particular focus on the Shelter Cluster during the Haiti earthquake response Using strategy documents, the Office for the Coordination of Humanitarian Affairs (OCHA)’s Financial Tracking Service and minutes of the Strategic Advisory Group (SAG) of the Haiti Shelter Cluster (the coordination structure for shelter agencies), we ask whether different decisions could have been made in early 2010. The article seeks to determine whether more questions could have been asked about the assumptions underpinning the response; learn from the Shelter Cluster’s strategic thinking in Haiti – one of the only published accounts of this process; offer a method of analysis for other clusters or responses; and illustrate the larger problems with the way the cluster system (and many other bureaucratic structures) has been conceived. It reflects on our roles without overinflating the importance of the international contribution, attributing blame, singling out the Shelter Cluster or dwelling on Haiti as an exceptional or complex context. These findings are not peculiar either to the Shelter Cluster in Haiti or the Global Shelter Cluster but are illustrative of larger problems with the way the Cluster System (and many other bureaucratic structures) has been conceived around the world.

Assumptions and money

Within days of the earthquake, the international community had conducted rough needs assessments. The displaced population was estimated at 70% of the 1.1 million people affected, against only 10% who remained ‘non-displaced on damaged homesteads’. Shelter Cluster, ‘Guiding Principles for the Emergency and Transitional Shelter and Settlement Strategy in Support of the Government of Haiti V5’, 28 January 2010.  The displaced were then designated as the priority group to receive assistance, even though the Shelter Sector Response Plan acknowledged that many had been displaced only short distances from their homes. The response plan categorised assistance in terms of individual, displaced households in urban settings, planned or self-settled sites (emergency and transitional shelter kits) or rural settings (host support).

crawford-figureThe Haiti Flash Appeal reflected this analysis in its allocation of resources to clusters (see figure). Funding for Camp Coordination and Camp Management (CCCM) outstripped other clusters and was awarded twice what had been requested. Three UN-Habitat proposals designed to strengthen Haitian capacity to coordinate urban response at national, municipal and neighbourhood levels were allocated only a fraction of the small amount requested and much less than any of the international clusters. The decision-makers and criteria involved in the allocation of this funding are not clear but, whatever the basis, this set in motion a coordination structure – around which people would then be gathered to report, analyse and strategise – that favoured the separate treatment of displaced and non-displaced groups, single-sector interventions and responses for individuals rather than groups.

Clusters and policies

There were thousands of affected and displaced people and we know that displacement contributes to vulnerability after disasters, but using ‘displacement’ as the priority category not only shaped coordination structures but also masked the complex and rapidly changing processes by which people sought shelter. The real-time evaluation of the response conducted by Groupe URD and GPPI three months after the earthquake noted that the population before the earthquake had been ‘highly mobile’. After the earthquake, people moved rapidly ‘within the city as well as between the city and rural areas’ and settled outdoors in fear of aftershocks. Francois Grunewald and Andrea Binder, Inter_agency Real_time Evaluation in Haiti: 3 Months After the Earthquake, Groupe URD and GPPI, 31 August 2010.  There was daily anecdotal evidence of larger camps spontaneously consolidating, secondary migration to these settlements, dispersal, return and eviction and people drifting back to Port-au-Prince after initially fleeing to rural areas.

As the categories of displaced and non-displaced were reinforced through the cluster structure, UN-Habitat – and other agencies supporting ‘return to neighbourhoods’ – recognised that circumscribing displacement camps as the unit of analysis meant that only camps and the people in them would be enumerated. The agency pushed for a conceptual shift – ‘neighbourhoods not camps’ – in February 2010, and this was partially reflected in the Humanitarian Country Team (HCT)’s safer shelter strategy. This did not offer any analysis of neighbourhoods, but it did link the exit from camps to spaces outside, promoting assistance for return, relocation to province of origin or relocation to a planned site. Government of Haiti and Interim Haiti Recovery Commission, Neighbourhood Return and Housing Reconstruction Framework; Stratégie de retour et de relocalisation, officially endorsed March 2011; Government of Haiti, Politique nationale du logement, de l’habitat et du développement urbain, April 2012.  These options appeared in early drafts of the national housing framework of October 2010, but were not translated into a strategic spatial approach connecting people to real places in the city until August 2011, when the Haitian government produced its 16:6 plan to decant people from six camps to 16 neighbourhoods in Port-au-Prince. Haiti Shelter Cluster,

Plans and standards

Prioritising support for the displaced but mandated to operate outside camps, the Haiti Shelter Cluster proposed a transitional shelter kit: a lightweight, notionally mobile asset (reacting to displacement and insecure tenure) but substantial in cost, size and durability (anticipating slow reconstruction). Overwhelmed procurement, logistics and installation teams were then organised to deliver this, reinforcing the tendency to overlook the original plan, which had included information campaigns, technical advice, cash, vouchers, materials distribution and rubble clearance (for displaced people); self-help, phased materials distribution and technical advice (for nondisplaced owners); and relocation assistance, rent and credit (for non-displaced renters).

Coordination via Technical Working Groups (TWGs), drawing on procurement and logistics staff from these teams, quickly became a process of harmonising technical standards, particularly square metres of Covered Living Space (Sphere SS&NFI Standard 3), unit costs and material specifications. This closed off opportunities to:

  • support cheaper or bespoke solutions for people who had been ‘mobile’ before the disaster and ‘nondisplaced’ or ‘slightly displaced’ afterwards, for example cash or repair programmes;
  • consider settlement planning (Sphere SS&NFI Standard 2), the spatial context into which Covered Living Space would be inserted and whether the projected number of shelters could even be squeezed into Port-au-Prince; and
  • challenge the priority given to displaced people and the preoccupation with improving ‘“security of tenure”, rather than addressing any identified insecurity of tenure’. Simon Levine et al., Avoiding Reality: Land, Institutions and Humanitarian Action in Post-earthquake Haiti, HPG Working Paper (London: ODI, 2012).

Questioning via the Strategic Advisory Group (SAG) mechanism

With clusters established, the responsibility for coordinating shelter and land issues was split between the CCCM Cluster (inside camps), Shelter (outside camps), the Logement-Quartier working group (housing and neighbourhoods) in the Early Recovery Cluster and the Housing, Land and Property (HLP) working group.

The Shelter Cluster convened its own Strategic Advisory Group (SAG), but the minutes suggest that this group identified but could not react to three key signals:

  • Doubts about data (April). The Shelter Cluster wanted a number of beneficiaries to set targets for fundraising and delivery of T-Shelter kits, but SAG participants realised that they did not have these numbers. Available data had not been gathered to assess need. Estimates of housing damage gauged the cost of the disaster; registration of people in camps (the Displacement Tracking Matrix (DTM)) was used to plan for resourcing camps; the Ministry of Public Works, Transport and Communications (MTPTC) Building Safety Assessment categorised housing as safe or unsafe to enter, but did not diagnose failures or allocate resources according to damage. The SAG was missing a synthesis of these datasets, broad-brush analysis of neighbourhoods and housing processes and information about the capacities of households, builders and markets.
  • Deliberate disconnect from housing (May). SAG minutes stressed repeatedly that transitional shelter should mean many forms of support, not just kits, and had to be linked to plans for housing. Coordination between the Shelter Cluster and the working groups looking at housing was ad hoc, and when housing was raised by the SAG one donor warned the group, correctly but unhelpfully, that it was dangerously close to the limit of its mandate.
  • Framing strategic problems as technical problems (April–-August). Responding to concerns about the cost and appropriateness of T-Shelters and the slow pace at which they were arriving, participants proposed various solutions: communicating successes and lobbying the government to expedite access to land for T-Shelter settlements; adding a note to T-Shelter kit parameters to account for taxes and inflation; accepting a lower target number of shelters (a reduction of 10–20%); flexibility on the location of kits rather than not questioning what would happen in places where kits were a poor fit; and wrangles over the relative merits of plastic sheeting versus plywood.

This failure to react is likely to stem from a combination of path dependency, bias and organisational politics. See James Darcy et al., Use of Evidence in Humanitarian Decision Making: ACAPS Operational Learning Paper, January 2013.  Once the assumptions, the flow of money and organisations and plans were in place, it was almost impossible to reconceive problems, counter-scenarios or geographic and strategic priorities. The embedded cluster logic to service camps and deliver shelter meant that problems were framed as obstacles to these objectives: lack of land for shelters and lack of government capacity to make decisions about land or accelerate approvals and procedures. In terms of bias, questions and datagathering centred on what agencies were doing, rather than what Haitians were already doing for themselves, and meetings were inaccessible to Haitian decisionmakers. Without analysis of activities other than the distribution of T-Shelter kits (and their relative and geographic impact), it was difficult to consider other non-permanent solutions, such as reinforcement kits; the appropriateness of T-Shelter kits for non-displaced populations; what non-assisted people would do and when; and trade-offs between alternatives, such as repairs, in terms of pace, cost and appropriateness. The final factor was organisational politics: UN-Habitat has a global urban mandate, chairs the Global Protection Cluster’s HLP Sub-Working Group, established Logement- Quartier and took over the HLP working group in June and the Shelter Cluster in November 2010. This was an opportunity for joining up many of the fragmented discussions, policies and data but, according to the SAG minutes, Habitat did not consistently attend the SAG, was under-resourced and appeared to be marginalised by peer agencies, perhaps for resisting the coordination paradigms it had challenged since the Flash Appeal.


The chain of events set in motion by early assumptions, the allocation of funding for coordination and an unexamined cluster structure all made it difficult to change course. Early estimates showing very small numbers of non-displaced people disconnected the territory of the camps from the fabric of the city and shaped the priorities, organisation, plans and data collection that followed. The destruction of high-density urban housing and infrastructure is likely to give rise to degrees of displacement that vary by population group, distance and duration, so we should think about the implications that these ‘degrees’ have for the assumption that displacement is the overriding determinant of urban vulnerability. Discussion of ‘appropriate’, ‘acceptable’ or ‘existing’ standards, required by Sphere, will be pointless unless we synthesise and interpret evidence about different places, historical shelter processes, infrastructure, markets and livelihoods, and develop a spatial account of our own actions that goes beyond ‘who, what and where’ to ‘why, when and how’. Strategy is better devised, at any scale, when we show realistic, honest and public assumptions about what might happen, how fast and where. With time, this can move from identification of risks to deeper consideration of particular courses of action and how and with whom to weigh up associated risks.

In terms of the money, funding for coordination followed and fed the logic of target groups (displaced versus nondisplaced), functions (servicing camps, providing individual shelters), shelters rather than places and fragmentation of the cluster’s efforts at analysis and synthesis. This compounded the better-known problem of separate operational funding for emergency shelter and reconstruction. Splitting responsibility for displaced and non-displaced people and embedding this in coordination structures drove strategy, rather than strategy driving organisation, and even when the strategic focus switched to return, policy frameworks remained disconnected from a strategic spatial approach to the city and its interrelated neighbourhoods. There was no concerted effort to create a spatial account – a shared platform combining urban and neighbourhood data (from before the disaster) with emerging damage and needs assessments, international activities and data on Haitian recovery. Without this, coordination in support of the government and the Haitian people has not been as useful or as open to scrutiny as it could have been then (or now). One serious consequence is that agencies were able to comply with the cluster strategy and with their own accountability frameworks while delivering, in the aggregate, a response that was homogeneous in activity and patchy geographically. Even with the sector’s traditional list- and zone-based tracking tools, we should have been concerned that a lot of large agencies were implementing only one activity from the shelter strategy.

A preference for technical answers over strategic questions forestalled thinking about vulnerable places, geographic priorities or community capacities. Locked into this path, discussions were limited to asset-based space standards not places, with a focus on space inside shelters not places on the ground and individual shelter units, not settlements. Flexibility in implementation was understood only in terms of T-Shelter kits: changing the numbers delivered, locations, unit costs and timescales for delivery – but strategic plans in complex emergencies are about revising holistic scenarios, not fiddling with single, predetermined solutions. The strategic advisory process did not call for resources to synthesise existing evidence and ignored the learning from successful responses elsewhere. Data collection followed the cluster logic and could not flesh out or expose our assumptions: when we made maps, we plotted only our own islands of activity; when we photographed shelters, we did not zoom out or pan to their context and setting.

The Shelter Cluster was not working in a vacuum. Some of the limits on mandate and failures to think about the broader picture were determined by other agencies and clusters, which themselves had little capacity for (and no contemporaneous account of ) strategic thinking. Humanitarian decision-making often happens in a context of scant evidence and overwhelming data. Leaders are rarely committing to a single solution but are rather signing up to a scenario, based on a bundle of implicit and interconnected assumptions about the best thing to do. But those assumptions need to be reviewed and this level of coordination – stitching back together the plans, synthesising the disparate data, building a shared spatial account of activities – did not seem to be happening anywhere in the wider humanitarian system.

Kate Crawford (Centre for Urban Sustainability and Resilience, University College London (UCL)) was previously Shelter Field Advisor for CARE International. Jim Kennedy was CARE Haiti’s Shelter Coordinator. Alison Killing is an urbanist. The opinions in this paper are those of the authors and do not represent the opinions of CARE or any other organisations mentioned.


Comments are available for logged in members only.